Article Text

Download PDFPDF

Original article
Fungal dysbiosis in cirrhosis
  1. Jasmohan S Bajaj1,
  2. Eric J Liu2,
  3. Raffi Kheradman2,
  4. Andrew Fagan1,
  5. Douglas M Heuman1,
  6. Melanie White1,
  7. Edith A Gavis1,
  8. Phillip Hylemon1,
  9. Masoumeh Sikaroodi1,
  10. Patrick M Gillevet2
  1. 1 Virginia Commonwealth University and McGuire Veterans Affairs Medical Center, Richmond, Virginia, USA
  2. 2 Microbiome Analysis Center, George Mason University, Manassas, Virginia, USA
  1. Correspondence to Dr Jasmohan S Bajaj, Division of Gastroenterology, Hepatology and Nutrition, Virginia Commonwealth University and McGuire VAMC, Richmond, VA 23221, USA; jasmohan{at}


Objective Cirrhotics have a high rate of infections, which are increasingly fungal or culture-negative in nature. While infected cirrhotics have bacterial dysbiosis, the role of fungi is unclear. We aimed to evaluate gut bacterial and fungal dysbiosis in cross-sectional and longitudinal analyses of outpatient and inpatient cirrhotics and prediction of hospitalisations.

Methods Cross-sectional: Age-matched controls, outpatients (with/without antibiotics) and hospitalised uninfected, culture-negative and culture-positive cirrhotics were included and followed for 90 days. Longitudinal: Three studies were conducted: (1) cirrhotics followed over 6 months, (2) outpatient cirrhotics administered antibiotics per standard of care for 5 days and (3) cirrhotics and controls administered omeprazole over 14 days. In all studies, stool bacterial/fungal profiles were analysed.

Results Cross-sectional: In 143 cirrhotics and 26 controls, bacterial and fungal diversities were significantly linked. Outpatients on antibiotics and patients with culture-positive infections had the lowest diversities. Bacterial and fungal correlations were complex in uninfected, outpatient and control groups but were markedly skewed in infected patients. 21% were admitted on 90-day follow-up. A lower Bacteroidetes/Ascomycota ratio was associated with lower hospitalisations. Longitudinal: Fungal and bacterial profiles were stable on follow-up (5 days and 6 months). After antibiotics, a significantly reduced bacterial and fungal diversity, higher Candida and lower autochthonous bacterial relative abundance were seen. After omeprazole, changes in bacterial diversity and composition were seen but fungal metrics remained stable.

Conclusion There is a significant fungal dysbiosis in cirrhosis, which changes differentially with antibiotics and proton pump inhibitor use, but is otherwise stable over time. A combined bacterial–fungal dysbiosis metric, Bacteroidetes/Ascomycota ratio, can independently predict 90-day hospitalisations in patients with cirrhosis.

Clinical trial number NCT01458990.


Statistics from

Significance of this study

What is already known on this subject?

  • Patients with cirrhosis have a high rate of morbidity and mortality from infections.

  • Most of these infections are presumed to be bacterial but culture-negative and fungal infections are on the rise due to extensive antibiotic use.

  • The interaction between fungi and bacteria and the systematic effect of antibiotics and proton pump inhibitors in this population are unclear.

What are the new findings?

  • A novel ratio, Bacteroidetes/Ascomycota, can predict 90-day hospitalisations in patients with cirrhosis independent of cirrhosis severity, hepatic encephalopathy and available clinical biomarkers.

  • Fungal diversity parallels bacterial values, is stable over time and is the most affected in cirrhotic outpatients on antibiotics and those with infections compared with uninfected inpatients and healthy controls.

  • Diversity in inpatients remains worse compared with outpatients on antibiotics.

  • Correlations between fungi and bacteria in culture-negative and culture-positive infected cirrhotics are skewed and simple compared with complex interactions in uninfected inpatients, outpatients and healthy controls.

  • In outpatients with cirrhosis who are initiated on broad-spectrum antibiotic therapy, there is a collapse of fungal and bacterial diversity but only bacterial diversity is affected with proton pump inhibitor use.

How might it impact on clinical practice in the foreseeable future?

  • Fungal and bacterial dysbiosis independently impact the hospitalisation risk and therefore both need to be considered when studying gut-based therapies to improve these outcomes.

  • Reduction in fungal diversity underlies antibiotic (rifaximin and spontaneous bacterial peritonitis prophylaxis) and culture-negative infections and a low threshold for antifungal therapy in these patients could be needed.

  • Outpatient antibiotic use in patients with cirrhosis leads to fungal diversity reduction, which should inform antibiotic stewardship and greater awareness of potential fungal causes of subsequent infections in these patients.


Cirrhosis is a major cause of death and disability worldwide.1 One of the major reasons for this mortality is related to infections that often stem from the gut.2 Gut bacterial dysbiosis in cirrhosis has been associated with these infections, which can lead to organ failure and death.3 The dysbiosis, or an unfavourable microbiome, is characterised by a lower autochthonous bacterial taxa (Ruminococcaceae, Lachnospiraceae, Clostridiales Cluster XIV).4–6 With the overuse of antibiotics in cirrhosis, culture-negative and fungal infections are being increasingly recognised.7 8 Cirrhotics have a high likelihood of developing infections because of altered immune response, multiple hospitalisations and instrumentations, use of proton pump inhibitors (PPIs) and the overuse of antibiotics for treatment and prophylaxis of infections.2 While most infections are presumed to be bacterial, culture-negative and fungal infections are being increasingly recognised.8 These have the potential to change strategies for further treatment and prophylaxis of infections.7 However, the role of gut fungal dysbiosis needs to be clarified.

We aimed to determine the presence of bacterial and fungal dysbiosis in patients with cirrhosis who have an infection (culture-positive and culture-negative) and on antibiotics compared with those without infections on antibiotics and healthy controls. We also aimed to determine its stability over time and study the effect of PPI and broad-spectrum antibiotic use.


Cross-sectional study

We recruited 143 patients with cirrhosis (77 outpatients, 66 inpatients) and 26 controls (table 1).

Table 1

Demographics and cirrhosis characteristics of subjects in the cross-sectional study

There were 47 infected patients, 22 culture-negative and 25 with culture-positive infections (see online supplementary table 1). Culture-positive infections were urinary tract infections (n=8, Escherichia coli 3, Enterococcus 2, Candida 3), Clostridium difficile (n=6), spontaneous bacterial peritonitis (SBP, n=6, Klebsiella 2, E. coli 2, Candida 2), SB empyema (n=1, Staphylococcus), bacteraemia (n=3,  Staphylococcus aureus n=1, Candida n=2) and one joint infection with S. aureus. Of the 22 culture-negative infections, most (n=17) had SBP, three had pneumonia and two had cellulitis. Uninfected Patients were hospitalised for ascites (n=13) and liver transplant evaluation (n=6). Second-generation cephalosporins were used in 21 patients and fluoroquinolones in the rest of the infected patients for median one dose prior. The patients who had Candida infections were sampled before antifungal therapies were initiated. 

Supplementary material 1

Bacterial profiles in those with higher fungal diversity

We defined low fungal diversity as a Shannon index of <1, 38% of subjects had low fungi compared with others. These patients had a lower relative abundance of Streptococcaceae, Clostridiales Cluster XIV and Bacillaceae with higher Enterococcaceae and Clostridiaceae.

Fungal and bacterial diversity changes with setting

Controls had the highest fungal and bacterial diversity and inpatients the lowest indices (figure 1A). There was a negative correlation between MELD score and fungal (r=−0.4, p=0.002) and bacterial (r=−0.43, p<0.001) Shannon diversity indices. There was also a significant correlation between fungal and bacterial diversity (figure 1B). Since patients on rifaximin and SBP prophylaxis had a higher model for end-stage liver disease (MELD) score than the rest (see online supplementary tables 2 and 3), therefore to define the role of absorbable and non-absorbable antibiotics, we then compared SBP prophylaxis versus rifaximin without prophylaxis and found these groups had a similar median MELD score (rifaximin 20 vs SBPP 19) without any change in fungal (rifaximin 0.77±0.77 vs SBPP 0.83±0.60) or bacterial (rifaximin 1.32±0.51 vs SBPP 1.3±0.5) diversity. Therefore, we combined the SBP prophylaxis and rifaximin groups as ‘outpatient antibiotics’.

Figure 1

Cross-sectional study diversity. (A) Shannon diversity in fungal and bacterial taxa was the highest in healthy controls and significantly lowest in inpatients. Data are presented as median and 95% CI with p values based on Kruskal-Wallis test. Ctrl, controls; Inpt, inpatients with cirrhosis; outpt, outpatients with cirrhosis. (B) Fungal diversity was significantly correlated with bacterial diversity. Black circles, infected patients with cirrhosis; red squares, uninfected patients with cirrhosis; green diamonds, healthy controls. (C) Shannon diversity indices for both fungi and bacteria were the highest in healthy controls and lowest in inpatients. Data are presented as median and 95% CI with p-values based on Kruskal-Wallis test. Ctr, controls; Cirr, outpatients with cirrhosis not on rifaximin, lactulose or SBP prophylaxis; Cneg, culture-negative infections; Cpos, culture-positive infections; Lac, outpatients with cirrhosis on lactulose; R+SB, outpatients with cirrhosis on rifaximin and/or SBP prophylaxis; Uninf, uninfected inpatients with cirrhosis.

Bacterial diversity in outpatient cirrhotics was lowest in those on antibiotics and within inpatients in patients with culture-positive infections (figure 1C). When outpatients on antibiotics were compared with patients with culture-positive infections, there was lower fungal and bacterial diversity in the patients with culture-positive infections (figure 1C). Endotoxin was correlated with MELD score (r=0.5, p=0.03) but even though inpatients had a higher endotoxin (table 1), there was no significant correlation with fungal diversity (r=0.1, p=0.73), while the bacterial diversity was marginally related (r=−0.4, p=0.05). There was no impact of diabetes on bacterial (yes 1.51±0.41 vs no 1.44±0.48, p=0.45) or fungal (yes 1.18±0.75 vs no 1.02±0.78, p=0.47) diversity indices.

Individual bacterial and fungal phyla are linked

The major fungal phyla, Ascomycota and Basidiomycota, were studied in relation to Bacteroidetes, Firmicutes and Proteobacteria in the entire cohort. As expected, there was a significant linkage between MELD score and Proteobacteria (r=0.4, p<0.0001) and Bacteroidetes (r=−0.4, p=0.014) but neither of the major fungal phyla nor Firmicutes were significantly linked to the MELD score. Ascomycota were negatively correlated with Bacteroidetes (r=−0.2, p=0.03). However, the Basidiomycota/Ascomycota ratio was positively correlated with MELD (r=0.3, p=0.05). Proteobacteria were negatively correlated (r=−0.3, p<0.001) with Firmicutes and to Bacteroidetes (p=−0.5, p<0.001). Bacteroidetes were also negatively linked with Firmicutes (r=−0.7, p<0.001).

Bacterial and fungal profiles were different between the groups on LEFSe

Healthy controls had a significantly higher relative abundance of autochthonous bacterial taxa as well as diverse fungi, including members of Basidiomycota when compared with outpatient cirrhotics (figure 2A). Similar findings were seen when healthy controls were compared with inpatients, except with a significantly higher relative abundance of Candida and potentially pathogenic bacterial taxa such as Enterobacteriaceae and Enterococcaceae in inpatients (figure 2B). When inpatients were compared with outpatients, a higher relative abundance of Candida, Enterobacteriaceae and Enterococcaceae was found in inpatients with a lower autochthonous bacterial taxa and members of Basidiomycota.

Figure 2

Cross-sectional study using linear discriminant analysis effect size (LEFSe). Specific changes in bacterial and fungal relative abundance between the three groups using LEFSe, Cirrhosis_No_antibiotics: Uninfected cirrhotics, Cirrhosis_Antibiotics: infected patients with cirrhosis, Controls: healthy controls. (A) Comparison between outpatients with cirrhosis (green) and healthy controls (red), (B) comparison between all inpatients with cirrhosis (green) and healthy controls (red), (C) comparison between outpatients with cirrhosis (green) and inpatients with cirrhosis (red), (D) comparison between outpatients with cirrhosis on antibiotics (rifaximin and spontaneous bacterial peritonitis prophylaxis; red) compared with outpatients with cirrhosis not on these medications (green).

Outpatients on antibiotics were more likely to have higher potentially pathogenic taxa such as Pasteurellaceae, and several members of Ascomycota, including Candida (figure 2D). This trend continued broadly with more Candida and fungi in infected patients, especially in culture-negative infections compared with culture-positive infections (see online supplementary tables 4–7).

Fungal phyla and Basidiomycota/Ascomycota changes with cirrhosis severity and groups

This ratio has been used to define the fungal dysbiosis, and in the cross-sectional analysis, there was a significant reduction of this ratio in inpatients and lowest in culture-positive infections compared with the other groups (figure 3A,B). When this was studied with respect to infections and antibiotic use, the lowest was in the outpatients on antibiotics and in infected inpatients (figure 3C).

Figure 3

Cross-sectional study of fungal taxa. (A) Significantly lower Basidiomycota/Ascomycota ratio in inpatients compared with outpatients and healthy controls. Data are presented as median and 95% CI with p-values based on Kruskal-Wallis test. Ctrl, controls; Inpt, inpatients with cirrhosis; Outpt, outpatients with cirrhosis. (B) Significantly lower Basidiomycota/Ascomycota ratio in inpatients with culture-positive infections compared with culture-negative, uninfected and healthy controls. Data are presented as median and 95% CI with p-values based on Kruskal-Wallis test. Cneg, culture-negative infections; Cpos, culture-positive infections; Ctrl, controls; Uninf, uninfected inpatients with cirrhosis. (C) Basidiomycota/Ascomycota ratio was low and statistically similar between outpatients on antibiotics and inpatient groups, compared with outpatients not on antibiotics and healthy controls. Data are presented as median and 95% CI with p-values based on Kruskal-Wallis test. Ctrl, controls; Inp-Inf, infected inpatients with cirrhosis; Inp-Uninf, uninfected inpatients with cirrhosis; OutAb, outpatients with cirrhosis on rifaximin and/or SBP prophylaxis; OutNAb, outpatients with cirrhosis not on rifaximin or SBP prophylaxis.

Correlations between bacteria and fungi were skewed in infected patients with cirrhosis

There were rich correlations between fungi and bacteria in healthy controls and outpatients with cirrhosis. This complexity reduced in inpatients without infections and culture-negative infections and was the lowest and most skewed in the patients with culture-positive infections (figure 4A–E).

Figure 4

Correlation networks between fungi and bacteria. (A) Correlation network demonstrates a dense linkage between bacteria (red nodes) and fungi (green nodes) with negative (red lines) and positive linkages (blue lines) in healthy controls. (B) Similar to healthy controls, correlation network in outpatient cirrhotics demonstrates a dense linkage between bacteria (red nodes) and fungi (green nodes) with negative (red lines) and positive linkages (blue lines). (C) Lower complexity in uninfected inpatient cirrhotics is seen between bacteria (red nodes) and fungi (green nodes) with negative (red lines) and positive linkages (blue lines). (D) Very low complexity in culture-negative infected inpatient cirrhotics is seen between bacteria (red nodes) and fungi (green nodes) with negative (red lines) and positive linkages (blue lines). (E) The lowest complexity in culture-positive infected inpatient cirrhotics is seen between bacteria (red nodes) and fungi (green nodes) with negative (red lines) and positive linkages (blue lines).

The novel Bacteroidetes/Ascomycota ratio was able to predict 90-day hospitalisations

Of the 143 patients, 90-day data were available for 123 patients because 12 died without getting readmitted and eight were lost to follow-up. Of these 123, 27 patients were admitted a median of 39 (IQR 14–71 days) postenrolment. All patients were admitted for liver-related reasons (hepatic encephalopathy (HE) n=14, infections n=6, anasarca n=4, hyponatraemia n=3, gastrointestinal (GI) bleeding n=2), of which four patients had multiple reasons related to HE, infections and GI bleeding. Inpatient status, infection, endotoxin levels, HE and specific microbial phyla were different at enrolment between those who were admitted compared with the rest (table 2). On univariate analysis, inpatient status, infection, HE, relative abundance of Bacteroidetes, Proteobacteria and Ascomycota had p<0.10. On multivariable analysis, HE (OR 3.5, CI 1.14 to 10.8, p=0.02) and Ascomycota relative abundance (OR 2.6, CI 1.02 to 8.8, p=0.04) were predictive, while Bacteroidetes relative abundance was associated with lower hospitalisations (OR 0.16, CI 0.02 to 0.98, p=0.05). We then calculated a Bacteroidetes/Ascomycota ratio, which then was again protective against 90-day hospitalisations (OR 0.10, CI 0.01 to 0.94, p=0.04) independent of HE (OR 5.3, CI 1.34 to 20.6, p=0.01). On Kruskal-Wallis, this median ratio was significantly lower in those who were hospitalised compared with those who remained free of hospitalisations (0.01 IQR 0.39 vs 0.47 IQR 0.54).

Table 2

Characteristics of patients who were admitted within 90 days

Prospective studies

Natural history study

There was no change in MELD score (8.8±2.9 vs 9.1±3.1, p=0.39), development of HE or other complications, or hospitalisations between the visits which were 6±1 months apart (table 3). On linear discriminant analysis effect size (LEFSe), there were no changes in individual bacterial or fungal taxa. This was also reflected in stable bacterial (visit-1, 1.54±0.40 vs visit-2, 1.43±0.37, p=0.38) or fungal (visit-1, 1.07±0.56 vs visit-2 1.31±0.56, p=0.11) diversity.

Table 3

Demographics and characteristics of subjects in the three longitudinal studies

Pre-PPI/post-PPI use 

All subjects were able to tolerate the omeprazole therapy as reported.9 There was no change in MELD score (8.1±2.7 vs 7.9±3.5, p=0.41) or new complications in the patients with cirrhosis during the trial (table 3). There were no changes in fungal taxa on LEFSe but as previously reported a significant change in bacteria reflecting oral microbial presence in the stool postomeprazole was seen (see online supplementary table 8).9 10 This was accompanied by a reduction in bacterial diversity in both controls (pre 1.88±0.16 vs post 1.63±0.22, p=0.004) and patients with cirrhosis (pre 1.89±0.25 vs post 1.54±0.34, p=0.05), while fungal diversity remained statistically unchanged in both groups (control pre1.92±0.48 vs post 1.73±0.61, p=0.46; cirrhosis pre 1.64±0.34 vs post 1.45±0.38, p=0.39).


Fourteen cirrhotics with hepatic encephalopathy controlled on lactulose and rifaximin were included; seven were administered the antibiotics, while the remaining were observed without it. No signs of infection were noted and the groups were balanced with respect to cirrhosis severity (table 3).

Fungal and bacterial diversity reduced and taxa changed after antibiotic use

Diversity of fungi and bacteria was statistically similar between groups at baseline but significantly reduced after antibiotic therapy (figure 5Atable 3). The Shannon diversity index remained statistically similar in those not administered antibiotics. Specifically on LEFSe, those given antibiotics had a significantly lower relative abundance of autochthonous bacterial taxa and of certain fungi compared with their baseline (figure 5B). In patients with cirrhosis who were not administered antibiotics, there were comparatively minimal changes in bacterial and fungal taxa between baseline and day 5. Compared with baseline, the antibiotics groups had a significant decrease in the relative abundance of autochthonous bacterial taxa and decrease in fungal taxa belonging to Sclerodermataceae and Dothideomycetes (figure 5C). After antibiotics, these fungi were not found and there was an increase in the relative abundance of taxa belonging to Chloroplast, Methylobacterium and Streptococcaceae. No significant LEFSe changes between baseline and day 5 of the no-antibiotic group were identified. At day 5 compared with the no-antibiotics group, there was a significant increase in the relative abundance of Candida in the antibiotics group, while other fungal taxa significantly decreased including Saccharomyces boulardii. The antibiotic group also had a significantly lower relative abundance of Bifidobacterium and Roseburia and higher relative abundance of Chloroplast and Methylobacterium (figure 5C).

Figure 5

Preoutpatient and postoutpatient antibiotics. (A) Shannon diversity indices for both fungi and bacteria were the lowest in patients with cirrhosis who received antibiotics compared with their baseline and both time periods of the no-antibiotics group. Data are presented as median and 95% CI with p-values based on Kruskal-Wallis test. Abx, antibiotics; Base, baseline; NAbx, no antibiotics; Post, after 5 days. (B) Specific changes in bacterial and fungal relative abundance after day 5 (green bars) compared with preantibiotic baseline (red bars) in the cirrhotic group receiving antibiotics using linear discriminant analysis effect size (LEFSe). (C) Cladogram showing a comparison at day 5 of the cirrhotics who received antibiotics (red) compared with those who did not (green) using LEFSe.


With the increasing prevalence of infections as a cause of organ failure and death in cirrhosis, there is a need to develop appropriate means to define their evolving microbiology.8 11 The study defines the important role of the gut mycobiome in outpatients and inpatients with cirrhosis and their evolving role in the prediction of future hospitalisations.

Our study spanned the entire spectrum of cirrhosis from compensated outpatients through patients who were hospitalised with and without infections. The results point towards a correlation between diversity in the fungal and bacterial taxa with worsening of the diversity with more advanced stages of disease. Specific bacterial phyla increase with worsening cirrhosis, that is, Proteobacteria as expected while the worsening MELD score was correlated negatively in the Basidiomycota/Ascomycota ratio.

On individual taxa analysis using LEFSe, this diversity was represented by a greater relative abundance of fungal taxa and autochthonous bacteria and lower Candida and potentially pathogenic bacteria such as Enterococcaceae in controls compared with inpatients or outpatients with cirrhosis. Similarly, autochthonous bacterial taxa and a greater variety of fungi spanning all phyla were seen in outpatients compared with a greater relative abundance of Candida and potentially pathogenic bacterial taxa in inpatients.

Within outpatients, the lowest bacterial and fungal diversity was found in those on antibiotics, which was also seen when broad-spectrum antibiotics were administered longitudinally. This bacterial diversity collapse with antibiotics cross-sectionally and longitudinally was accompanied by an increase in relative abundance of Ascomycota components such as Candida. These findings in the longitudinal cohort are even more striking given that these were patients already on rifaximin and lactulose. Despite this skewed gut milieu, broad-spectrum antibiotics were able to further disrupt the bacterial and fungal population, indicating the pervasive, additive impact of antibiotics that are used rampantly and commonly in this group.

Antibiotics are one of the major risk factors for emergence of fungi,12 13 which in our study was evaluated comparing uninfected patients with those with culture-negative and culture-positive infections and outpatients on antibiotics.14 While infected and uninfected inpatients were similar on MELD score and diversity as a whole, patients with culture-positive infections demonstrated a lower diversity, lower Basidiomycota/Ascomycota and relative abundance of fungi on LEFSe compared with patients with culture-negative infections. This diversity was also lower than in antibiotic-using outpatients. These findings point towards an inherent gut milieu in patients with culture-positive infections, a proportion of which had Candida infections, which is independent of antibiotics and MELD score and could be a target of modulation with strategies other than further antibiotic therapy.

Interestingly, as opposed to antibiotic therapy, PPI therapy did not significantly affect fungal diversity in controls or cirrhotics, in contrast to bacterial diversity. While the mechanism is not clear, it could be due to the reduction in total bacterial abundance seen with antibiotics, which is not usually seen with PPIs.15 Being saprophytic, fungi are dependent on bacteria for nutrition, and it is possible that substitution of the usual faecal microbiota with oral microbiota in the case of PPI use could still maintain their food source while a significant reduction in the entire population abundance, as occurs with antibiotic use could secondarily impact fungi. This nuanced interaction is also represented by the lack of correlation between fungal diversity and endotoxin and between fungal phyla and MELD directly. Therefore, fungal presence is not increased directly by the usual gram-negative bacterial taxa increase that accompanies advancing cirrhosis but could be modulated by other factors. Moreover, the complex correlations between fungi and bacteria that existed in uninfected inpatients, outpatients and controls were reduced to a skewed linkage pattern in infected patients, both culture-positive and culture-negative. This was similar to patterns seen in inflammatory bowel disease, although we did not find significant alterations in Saccharomyces cerevisiae , lending the findings specificity for cirrhosis.16 Interactions between these two kingdoms is complex due to multiple nutritional, quorum-sensing and competitive interactions17 18 and based on our study, is differentially affected by antibiotics, acid suppression or cirrhosis severity.

Interestingly, there were no changes in the bacterial or fungal diversity or in composition on LEFSe in outpatients with cirrhosis followed either short-term (advanced group randomised to no-antibiotics) or long-term (on 6-month follow-up). This extends prior studies of bacterial diversity into the fungal realm and demonstrates the underlying resilience of this composition, provided the underlying cirrhosis course is stable.3 The resilience was also reflected by the non-significant impact of diabetes, whose effect may have been diluted by the already skewed microbiota in cirrhosis, as previously shown.19 20

The current clinical practice for treating signs and symptoms suggestive of infections (high WBC count, SIRS criteria, hepatic encephalopathy) in patients with cirrhosis even without an identifiable micro-organism is to initiate antibiotics to treat a presumed bacterial source.21 The current results demonstrate a reduction in fungal diversity with overabundance of Ascomycota in culture-negative and even in bacterial culture-positive infections. The most common fungal infections in cirrhosis, peritonitis, fungaemia and esophagitis are usually due to Ascomycota members and have a presumed gut origin.8 13 18 These infections, when clinically confirmed using current culture techniques, carry a devastating prognosis.22 A dysbiotic mycobiome is characterised by an altered Basidiomycota/Ascomycota and higher Ascomycota, which was indeed also observed in advanced and infected cirrhosis could be exacerbated by this rampant antibiotic use.13 22 Our results also show that this higher relative abundance of Ascomycota could also predict hospitalisations within 90 days independent of bacterial components of the microbiome and clinical biomarkers. This was in contrast to the potentially protective role of Bacteroidetes that replicates prior experience.20 Interestingly, the ratio of Bacteroidetes/Ascomycota rather than Basidiomycota/Ascomycota was predictive of hospitalisations, underlining the finding that both mycobiome and bacterial microbiome needs to be taken into consideration to potentially improve prognostication.

The study is limited due to lack of fungal isolation from some samples, which is often seen in mycobiome studies.23 This could be due to a lower absolute abundance of fungi in the compared with bacteria and therefore, in a defined DNA sample, one is subsampling fewer molecules. We also had few confirmed fungal infections that did not give us the sample size to separate those from culture-positive bacterial infections. The study does not clearly demonstrate specificity for cirrhosis versus antibiotic use, although we did not find changes in Saccharomyces as shown in IBD and found that fungal relative abundance predicted hospitalisations.

Cumulatively, our results demonstrate that there are systematic reductions in fungal diversity which parallel bacterial diversity in outpatients and inpatients with cirrhosis. This dysbiosis changes differentially with antibiotics and PPI use, but is otherwise stable over time. A combined bacterial–fungal dysbiosis metric, Bacteroidetes/Ascomycota, can independently predict 90-day hospitalisations in patients with cirrhosis. These findings demonstrate that fungi are major modulators of the overall gut microbial dysbiosis in cirrhosis and further studies to determine if reducing the relative abundance of Ascomycota can prevent hospitalisations or development of fungal infections are needed.


A cross-sectional and three longitudinal studies were performed. The protocols were approved by IRBs at the VCU and Richmond VA Medical Centers, and all subjects gave written informed consent for the study:

Cross-sectional study 

We enrolled healthy controls that were age-matched with two sets of patients with cirrhosis, outpatients and inpatients. Healthy controls were outpatients without any chronic diseases or medications. We included patients with cirrhosis who were diagnosed using liver biopsy, evidence of frank decompensation or radiological features of cirrhosis who were between 21 and 75 years. Patients who were unable to give consent or provide stool within 48 hours of admission, with alcohol misuse, alcoholic hepatitis, on antifungal medications and with gastrointestinal bleeding episodes within 6 weeks were excluded. We also excluded those undergoing HCV eradication or those on probiotics within the last 3 months. These eligibility criteria for cirrhosis and healthy controls were used for all cross-sectional and longitudinal studies.

All patients had medications and diabetes status recorded. Blood was drawn for endotoxin assay (Limulus amebocyte lysate assay) and stool collected for microbiota analysis.

Outpatients with cirrhosis were recruited from the clinics after informed consent. Inpatients with cirrhosis consisted of three groups: (1) uninfected, (2) culture-negative and (3) culture-positive infections. For the culture-negative group, we included all infected patients whose routine bacterial and fungal cultures were negative for any organism but had signs of infection. These included >250 polymorphonuclear cells on ascites fluid with negative culture for SBP, specific radiological features of pneumonia without positive sputum or blood culture and so on.8 All these had received at least one dose of antibiotics per standard of care before stool collection. The patients with culture-positive infections had one or more bacterial or fungal organisms isolated as the cause. However, the stool collection had taken place before that determination was made. In the uninfected group, the included patients fit the same criteria for hospitalisation without an infection either suspected or documented and without current absorbable antibiotic use for presumed infections.


All patients were followed up for 90 days or until death or liver transplant. Non-elective hospitalisations 90 days postenrolment were studied. Clinical and microbiota (fungal and bacterial) parameters were compared. A multivariable backwards logistic regression model was created by including clinical and microbiota (fungal and bacterial) variables that were p<0.10 on univariate analysis with 90-day hospitalisation as the outcome.

Longitudinal studies

Natural history study

We enrolled outpatients with cirrhosis who gave stool samples at least 6 months apart with stability of the cirrhosis course between visits.


From a prior trial in which we administered 40 mg of omeprazole daily for 14 days in compensated outpatients with cirrhosis and healthy controls, we analysed the impact of PPI on fungal parameters.9

Antibiotics compared with standard of care: A cohort of outpatients with prior hepatic encephalopathy on lactulose and rifaximin were recruited after written informed consent and were randomised into two groups using a random number generator; one group was administered 5 days of broad-spectrum antibiotics (metronidazole 500 mg three times a day, ciprofloxacin 500 mg twice a day and amoxicillin 500 mg twice a day), while the other group was followed without intervention for the same duration. Stool samples were collected at baseline and at day 5.

For all these studies, Shannon diversity indices for both bacterial and fungal taxa as well as LEFSe comparisons24 at baseline and the end of study were performed and compared. Detailed microbial methods are in given in online supplementary methods.



  • Contributors JSB conceptualised the study and was involved in all aspects. PMG was involved in the microbiota, bioinformatics and drafting and editing the paper. EJL, RK and MS were involved in microbiota and bioinformatics analysis. PH and DMH were involved in critical revision. MW, EAG and AF helped with subject enrolment.

  • Funding This work was partly supported by VA Merit Review CX10076 and McGuire Research Institute.

  • Competing interests No competing financial interest exists for any author.

  • Patient consent Obtained.

  • Ethics approval Virginia Commonwealth University and McGuire VAMC IRBs.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.